Inverse design of metal nanoparticles based on deep learning
Rui Wang, Chunlan Liu, Yong Wei, Ping Wu, Yudong Su, Zhi Zhang
Abstract
We proposed and demonstrated a novel inverse design method of metal nanoparticles based on deep learning. By selecting the position of the SPR resonance peak, we designed the size and storage status of the metal nanoparticles. Taking gold nanospheres as an example, we contrasted the inverse design method of gold nanoparticles based on least square method and neural network for error back propagation (BP) training. The calculation results indicated that the inverse design method based on deep learning is more adaptable and stable, and its minimum error reached 2.23 × 10−12. The proposed inverse design method based on deep learning can change the mathematical model of metal nanoparticles according to the actual demands, which is suitable for the inverse design of more materials, morphology and structure of nanoparticles. It provides a new idea for the design of nanoparticles.